A coalition of security professionals from Washington and Beijing have outlined a system of guardrails modeled on Cold War nuclear protocols to mitigate artificial intelligence risks. The proposal focuses on preventing an accidental war triggered by autonomous systems that might launch military cyber operations or disrupt nuclear command networks without human authorization. Such a malfunction could leave officials with only minutes to decide if they were targeted by a deliberate attack by the opposing government.

The recommended safeguards include a dedicated AI military hotline and the requirement for human approval for all nuclear launch decisions and cyberattacks. These proposals were developed during a security conference in China and precede scheduled bilateral talks and a meeting between Donald Trump and Xi Jinping next week. Neither government has formally signed onto the plan, though experts warn of the risks as automated systems interact at machine speed.